Inside Systems 03: You Read the Output. You Did Not Check It.

The real danger is not AI errors. It is how quickly humans stop checking for them.